``what'' and ``where'' in auditory sensory processing: a

INTEGRATIVE NEUROSCIENCEORIGINAL RESEARCH ARTICLE

published: 22 June 2011doi: 10.3389/fnint.2011.00023

“What” and “Where” in auditory sensory processing: ahigh-density electrical mapping study of distinct neuralprocesses underlying sound object recognition andsound localization

Victoria M. Leavitt 1,2†, Sophie Molholm1,3,4, Manuel Gomez-Ramirez1† and John J. Foxe1,2,3,4*

1 The Cognitive Neurophysiology Laboratory, Nathan S. Kline Institute for Psychiatric Research, Orangeburg, New York, NY, USA2 Program in Neuropsychology, Department of Psychology, Queens College of the City University of New York, Flushing, NY, USA3 The Cognitive Neurophysiology Laboratory, Children’s Evaluation and Rehabilitation Center, Department of Pediatrics, Albert Einstein College of Medicine,

Bronx, NY, USA4 Department of Neuroscience, Albert Einstein College of Medicine, Bronx, NY, USA

Edited by:

Patricia M. Di Lorenzo, BinghamtonUniversity, USA

Reviewed by:

Klaus Mathiak, RWTH AachenUniversity, GermanyRichard Pastore, Purdue University,USA

*Correspondence:

John J. Foxe, Departments ofPediatrics and Neuroscience, AlbertEinstein College of Medicine, VanEtten Building – Wing 1C, 1225 MorrisPark Avenue, Bronx, NY 10461, USA.e-mail: [email protected]†Present address:

Victoria M. Leavitt, Neuropsychologyand Neuroscience Laboratory, KesslerFoundation Research Center, WestOrange, NJ, USA;Manuel Gomez-Ramirez, ZanvylKrieger Mind/Brain Institute, JohnsHopkins University, Baltimore, MD,USA

Functionally distinct dorsal and ventral auditory pathways for sound localization (WHERE )and sound object recognition (WHAT ) have been described in non-human primates. A hand-ful of studies have explored differential processing within these streams in humans, withhighly inconsistent findings. Stimuli employed have included simple tones, noise bursts,and speech sounds, with simulated left–right spatial manipulations, and in some casesparticipants were not required to actively discriminate the stimuli. Our contention is thatthese paradigms were not well suited to dissociating processing within the two streams.Our aim here was to determine how early in processing we could find evidence for disso-ciable pathways using better titrated WHAT and WHERE task conditions.The use of morecompelling tasks should allow us to amplify differential processing within the dorsal andventral pathways. We employed high-density electrical mapping using a relatively large andenvironmentally realistic stimulus set (seven animal calls) delivered from seven free-fieldspatial locations; with stimulus configuration identical across the “WHERE” and “WHAT”tasks. Topographic analysis revealed distinct dorsal and ventral auditory processing net-works during the WHERE and WHAT tasks with the earliest point of divergence seenduring the N1 component of the auditory evoked response, beginning at approximately100 ms. While this difference occurred during the N1 timeframe, it was not a simple mod-ulation of N1 amplitude as it displayed a wholly different topographic distribution to that ofthe N1. Global dissimilarity measures using topographic modulation analysis confirmed thatthis difference between tasks was driven by a shift in the underlying generator configura-tion. Minimum-norm source reconstruction revealed distinct activations that correspondedwell with activity within putative dorsal and ventral auditory structures.

Keywords: auditory, event-related potential, electrophysiology

INTRODUCTIONThere is growing evidence for functionally and anatomically dis-tinct auditory processing pathways. Findings in non-human pri-mates (Romanski et al., 1999; Rauschecker and Tian, 2000; Tianet al., 2001) and humans (Weeks et al., 1999; Clarke et al., 2000,2002; Anourova et al., 2001; Zatorre et al., 2002; Poremba et al.,2003) point to a WHAT/WHERE distinction analogous to thatseen in the visual system (Mishkin et al., 1982). These data sug-gest a system whereby auditory inputs are communicated dorsallyfrom primary auditory cortex for the extraction of informa-tion regarding spatial localization in parietal cortical regions (theWHERE pathway), and ventrally from primary auditory cortexalong medial and inferior temporal cortex for the processing ofthe specific object features of the signal such as spectral contentand temporal integration (the WHAT pathway).

Much of what we know about WHAT and WHERE audi-tory pathways comes from studies in non-human primates, whichsuggest that the earliest evidence for separable and functionallydistinct pathways can already be found in the medial geniculatenucleus (MGN) of the thalamus (Kaas and Hackett, 1999; Roman-ski et al., 1999; Rauschecker and Tian, 2000). Rauschecker andTian (2000) summarized work utilizing histochemical techniquesand anatomical tracer dyes to track the trajectory of neuronalsubpopulations from the MGN to primary auditory cortex (A1).Adjacent to A1 are a rostral area (R) and a caudomedial area(CM; Figure 1). While A1 and R have been shown to receiveinput from the ventral portion of the MGN, CM is the targetof dorsal, and medial divisions of MGN. The lateral belt areasreceiving input from A1 and R may represent the beginning of anauditory pattern or object stream, as the neurons here respond to

Frontiers in Integrative Neuroscience www.frontiersin.org June 2011 | Volume 5 | Article 23 | 1

brought to you by COREView metadata, citation and similar papers at core.ac.uk

provided by Frontiers - Publisher Connector

https://core.ac.uk/display/82841798?utm_source=pdf&utm_medium=banner&utm_campaign=pdf-decoration-v1

http://dx.doi.org/10.3389/fnint.2011.00023

mailto:[email protected]

www.frontiersin.org

http://www.frontiersin.org/integrative_neuroscience/10.3389/fnint.2011.00023/abstract

http://www.frontiersin.org/people/johnfoxe/983

http://www.frontiersin.org/Integrative_Neuroscience/

http://www.frontiersin.org/Integrative_Neuroscience/about

http://www.frontiersin.org/Integrative_Neuroscience/editorialboard


http://www.frontiersin.org/Integrative_Neuroscience/archive

Leavitt et al. Dorsal/ventral auditory processing in humans

FIGURE 1 | Anatomical regions of non-human primate (monkey) brain

involved in processing of spatial and non-spatial stimuli. (A) Lateralbelt region (in color) comprises A1, primary auditory cortex; CL,caudolateral belt region; AL, anterolateral belt region; ML, medial lateral beltregion. (B) The same region, enlarged to show the locations of the medialand lateral belt (yellow), primary auditory cortex (purple) and the parabeltcortex on the superior temporal gyrus. Adapted from Romanski et al.(1999), with permission.

species-specific vocalizations and other complex sounds (Roman-ski et al., 1999; Rauschecker and Tian, 2000). Conversely, neuronsin caudal regions CM and CL (caudomedial and caudolateralregions respectively) show a good degree of spatial selectivity (seeTian et al., 2001). Romanski et al. (1999) traced the streams fur-ther in rhesus monkeys, finding distinct targets for the ventral anddorsal pathways in non-spatial and spatial domains of prefrontalcortex, respectively. Similarly, human lesion studies have providedadditional support for the division of the auditory system into dor-sal and ventral pathways with clear functional dissociations seenfor patients with dorsal versus ventral lesions (Clarke et al., 2000,2002; Clarke and Thiran, 2004).

In humans, there is also mounting functional evidence forWHAT and WHERE divisions of the auditory system. Arnottet al. (2004) compiled a meta-analysis of neuroimaging studiesof ventral and dorsal auditory processing in humans. By review-ing 11 spatial studies and 27 non-spatial studies, these authorsconcluded that physiologically distinct areas for processing spa-tial and non-spatial features of sounds could be identified. Forthe studies classified as “spatial,” centers of activation clusteredaround the inferior parietal lobule, the temporal lobe posteriorto primary auditory cortex, and the superior frontal sulcus. Forthe non-spatial studies, centers of activation were seen clusteredaround inferior frontal cortex. While there was substantial over-lap in areas of activation, the clusters of activation correspondedwell with the notion that WHERE processing occurs in a dorsallypositioned processing stream and WHAT processing in a ventrallypositioned processing stream (see Figure 2). The studies includedin this analysis all utilized positron emission tomography (PET)and/or functional magnetic resonance imaging (fMRI). Thus far,however, there have been relatively few electroencephalographic(EEG)/magnetoencephalographic (MEG) investigations of pro-cessing within these pathways in the human auditory system.Therefore, while we have some idea of the anatomical localiza-tion of the processing of these distinct types of information, ourunderstanding of the temporal dynamics of these processes is stillquite sparse.

FIGURE 2 | Schematic representation of auditory pathways as

evidenced by a meta-analysis of 11 spatial (dorsal ) and 27 non-spatial

(ventral ) studies. Regions showing greatest activation in each study areindicated. Adapted from Arnott et al. (2004) with permission.

Furthermore, the paradigms that have been employed to tapinto functional capacities of auditory processing have variedgreatly. For instance, a large portion of the non-spatial studiesexamined in the Arnott meta-analysis used human speech sounds,and this may be problematic insofar as speech represents a highlyspecialized auditory information class for humans, with a spe-cific set of speech processing regions that may not necessarily lendthemselves easily to this dorsal/ventral distinction. It has also beenproposed that the dorsal pathway in humans not only processesauditory space, but also plays an important role in speech percep-tion (Belin and Zatorre, 2000), raising the issue of possible overlapof presumptive dorsal and ventral pathway functions.

Human EEG and MEG studies have certainly attempted to shedsome light on the timing of dorsal/ventral processing dissocia-tions, but the findings have been highly inconsistent in terms ofhow early in the processing hierarchy information begins to bepreferentially processed by a specific pathway. In the context of amatch-to-sample task for location (right and left) and frequency(220, 440, and 880 Hz) that varied by memory load, Anurova et al.(2003) found very late effects of task differences occurring approx-imately 400 ms after stimulation. In an fMRI/EEG study, Alainet al. (2001) also found late occurring task-related effects from300 ms onwards using a match-to-sample task. Stimuli in theirstudy were five pitch-varied noise bursts delivered via headphonesto five simulated spatial locations. In contrast to these findings oflate processing differences, Anourova et al. (2001) using simulta-neously recorded EEG and MEG, found peak latency differencesbetween WHAT and WHERE tasks at considerably earlier time-points during the so-called N1 component of the AEP (∼100 ms),but there were no amplitude differences seen. In that study, iden-tical blocks of tone stimuli were presented to subjects who wereinstructed to attend either to sound location or frequency. Stimuliwere simple tones of two frequencies (1000 and 1500 Hz) delivered


www.frontiersin.org




via headphones to two simulated locations (an interaural intensitydifference was applied; perceived location was either left or right).Utilizing MEG and fMRI and a match-to-sample task with speechstimuli, Ahveninen et al. (2006) found both amplitude and latencyeffects at ∼70–150 ms (during the N1 component). De Santis et al.(2007a) also argued for “WHAT/WHERE” related effects on theamplitude of the auditory N1, but it is important to point outthat in this study, subjects were involved in a passive listeningtask and did not explicitly perform spatial or object judgments.Stimuli were bandpass filtered noise bursts (250 or 500 Hz) deliv-ered to two perceived locations (interaural time difference wasused to simulate left and right locations). Although the aim ofthis study was for a setup where the stimuli used in the WHATand WHERE conditions were “identical,” this was not in fact thecase, and the design resulted in unique stimulus configurations foreach condition. For the WHAT condition, 80% of the stimuli wereof one frequency and 20% were of a second frequency, while thespatial position was equi-probable (50/50). A second counterbal-anced block of the same passively presented WHAT condition wasrun in which the 80/20 ratio was reversed, and for analysis, theauthors appropriately collapsed responses over both blocks. Ananalogous 80/20 design was used for the two blocks of the putativeWHERE condition, with the frequency aspect of the stimulatedstream varied equally (50/50). There appear to be two potentialissues with this design that complicate a straightforward interpre-tation of the results. First, the 80/20–50/50 ratios for one conditionwere flipped for the other and so the stimulation sequences werenot identical. Secondly, and more importantly, this design takes theform of a standard oddball paradigm, and would necessarily haveproduced the well-known mismatch negativity (MMN) to the low-probability stimuli (see e.g., Molholm et al., 2005; Näätänen et al.,2007). To circumvent this issue only the frequent tones were com-pared between conditions. Unfortunately, it has been shown thatthe first frequent“standard”tone following an infrequent“deviant”tone also receives differential processing under such designs, anda robust MMN is often seen to this presumed “standard” tone (seee.g., Nousak et al., 1996; Roeber et al., 2003). Upon careful reading,responses to these first standards appear to have been included inthe De Santis analysis (De Santis et al., 2007b). In all five of the pre-vious EEG/MEG studies that aimed to study the WHAT/WHEREdistinction in audition, stimuli were tones, noise bursts, or vowelsounds, and all utilized simulated three-dimensional (3D) envi-ronments delivered via headphones. None of these studies usedsounds delivered in the free field and one could certainly questionthe “objectness” of simple tones and noise bursts. There is also aclear discrepancy between the two studies that show only late dis-sociations (>300 ms: Alain et al., 2001; Anurova et al., 2003) andthose that suggest considerably earlier effects during the N1 pro-cessing period (∼100 ms: Anourova et al., 2001; Ahveninen et al.,2006; De Santis et al., 2007a).

A prior investigation of mid-latency auditory evoked responses(MLRs) from our group suggested that there was constitutivedivergence of activation into separable pathways in the context ofa passive listening task, a divergence that is already observable veryearly in processing during the auditory Pa component,which peaksat just 35 ms (Leavitt et al., 2007). At this latency, source-analysispointed to distinct responses with sources from more dorsal and

more ventral areas of auditory cortex. While that study was notdesigned to expressly test the functionality of the dorsal and ventralauditory pathways, automatic sensory activation of these pathwaysis likely an inherent result of receiving any auditory stimulationor performing any auditory task. By analogy, visual stimuli con-stitutively activate both the dorsal and ventral visual streams evenwhen no explicit object or spatial processing tasks are required(see, e.g., Schroeder et al., 1998). That is, both pathways are acti-vated by typical visual inputs, but it is possible to activate them inrelative isolation by manipulating stimulus properties (e.g., the useof isoluminant chromatic contrast; see Lalor et al., 2008). Indeed,in non-human primate studies, the animals are often anesthetizedduring recordings and yet both the dorsal and ventral pathwaysare activated (e.g., Schmolesky et al., 1998). Our aim here was todetermine just how early in auditory processing we could find evi-dence for dissociable pathways under specific WHAT and WHEREtask conditions. Our premise was that functionally distinct andecologically valid tasks would allow us to amplify any differen-tial processing within the dorsal and ventral pathways, and thatthere were likely earlier dissociations of WHAT/WHERE process-ing than previously reported (e.g., Alain et al., 2001; Anourovaet al., 2001; Anurova et al., 2003; Ahveninen et al., 2006; De Santiset al., 2007a). To that end, we employed animal calls as stimuli,avoiding any interference from language-specific areas that maybe activated in response to human speech sounds, as well as tak-ing advantage of the broadened extent of auditory cortex that isactivated by complex acoustic stimuli (as opposed to simple tones;see for example Rama et al., 2004). In using a larger set of com-plex animal calls and thereby attaching semantic meaning to ourstimuli, our intention was to maximize object-processing whileavoiding possible confounds induced by the use of speech stimuli.Similarly, by using free-field spatial stimuli delivered from a rel-atively large array of seven possible locations, we intended to getaway from simple left–right judgments over headphones that mayhave been less effective at invoking spatial mechanisms. Identicalstimuli were maintained for both tasks, and only the task itself wasvaried.

MATERIALS AND METHODSThe procedures for this study were approved by the Institu-tional Review Board of The Nathan S. Kline Institute for Psychi-atric Research/Rockland Psychiatric Center and the InstitutionalReview Board of The Graduate Center of the City University ofNew York.

SUBJECTSInformed consent was obtained from 12 (10 male) healthy con-trol subjects aged 23–56 years (mean = 33 ± 11.4). All reportednormal hearing. One subject was left-handed, as determined bythe Edinburgh Test for Handedness (Oldfield, 1971). None of theparticipants had current or prior neurological or psychiatric ill-nesses, nor were any currently taking psychotropic medications.All subjects were paid a modest fee for their participation.

STIMULIPrecisely the same stimuli were used for both the WHAT andWHERE conditions: seven animal sounds (cow, dog, duck, ele-phant, frog, pig, and sheep), adapted from Fabiani et al. (1996).


www.frontiersin.org




These sounds were of uniquely identifiable vocalizations. Theywere modified using Cool Edit Pro Version 2 such that each hada duration of 250 ms, and were presented over seven BlaupunktPCx352 90 watt 3.5′′ 2-way coaxial sound sources at a comfortablelistening level of approximately 80 dB sound pressure level. Inter-stimulus interval (ISI) was variable to avoid anticipatory effects;stimuli in both conditions were delivered every 2500–3500 ms.

PROCEDUREParticipants were seated in a comfortable chair in a dimly illumi-nated, sound-attenuated, electrically shielded (Braden ShieldingSystems) chamber and asked to keep head and eye movements toa minimum. Auditory stimuli were presented from seven spatiallocations via sound sources arranged in the participant’s medianplane along a frontal arc of ±90˚ (−90˚, −60˚, −30˚, 0, +30˚,+60˚, +90˚) and centered on the participant’s head (1.0 m dis-tance; see Figure 3). This arc was adjustable on a pulley systemsuch that it could be positioned level with the horizontal planeof each individual subject at head height. Sound sources wereconcealed behind a black curtain to prevent participants fromforming a visual representation of sound source positions. Theapparatus was precisely centered within the 2.5 m × 2.5 m squarechamber, equidistant from the left and right walls of the cham-ber. A 12′′ × 12′′ digitized Wacom graphics tablet was mounted infront of and aligned with the participant’s median sagittal planeof the trunk. It consisted of a stylus that the participant used todraw a freehand line toward the source of the sound as it wasperceived (during the localization task). The tablet was obscuredfrom view by a partition. There were two blocked conditions,which were counterbalanced for order across subjects. The twoconditions were equated for motor (response) requirements, tominimize motor confounds. In addition, the exact same stimuliwere used in both conditions, to minimize potential confounds.In order to insure sufficient power, we collected a very large num-ber of sweeps for each subject (upward of 2000 trials on averagein both conditions, an order of magnitude more than is typical,thereby ensuring very high SNR). The number of blocks varied

in the WHERE condition due to individual subject factors (i.e.,fatigue, breaks, time constraints). (1) WHERE task: the WHEREtask was to indicate the location that a sound came from; subjectscompleted a minimum of 13 and not more than 15 blocks. Eachblock contained 168 stimulus presentations (24 per sound source).Stimuli were animal calls, as in the WHAT task. Spatial presenta-tion of the stimuli followed a random order, as did the animal calls.Subjects responded by drawing a line on the graphic tablet out tothe radius in the direction of each animal call after it was delivered.The angle and trajectory of the line was recorded by the computerin Presentation 4.0. Then, the participant returned the stylus tothe midpoint of the tablet; the resting position. This position wasindicated by a depression in the tablet. (2) WHAT task: the WHATtask was to identify a target animal from seven randomly deliveredanimal calls. The target animal for each run was randomly selectedand randomized across blocks. There were 14 blocks. At the startof each block, the subject was given an auditory cue: “This is yourtarget. . .” followed by one of the seven randomly selected targetanimal calls. After that, all seven animal calls were randomly deliv-ered throughout the block in equal distributions, precisely as inthe WHERE task. In this task subjects were to respond to a desig-nated target with one response, and to all non-targets with anotherresponse. There were two response types, drawing either a line ora circle on the tablet. The pairing of response type to stimulustype (target or non-target) was counterbalanced across subjects.The target stimuli occurred on 15.6% of trials. Each animal servedas target in 2 of 14 runs, in randomized order both within andbetween subjects. To address a potential stimulus confound, analternate WHAT task, WHAT II, was run on a subset of subjects(6 of the 12 subjects). The concern was that in the WHERE task,each stimulus was processed by subjects as a target requiring aresponse. However, the WHAT task stimuli were targets and non-targets. Therefore, in WHAT II subjects were instructed: “Pleasesay the name of each sound you hear,” thereby equating all stimuliacross conditions as targets. (Note that while subjects were neverexplicitly told they would hear“animal”calls, so as not to bias themto semantic category, it was very clear that all subjects immediately

FIGURE 3 | Graphic representation of sound source locations. Seven sound sources were arranged in the participant’s median plane along a frontal arc of±90˚ (−90˚, −60˚, −30˚, 0, +30˚, +60˚, +90˚) and centered on the participant’s head (1.0 m distance).


www.frontiersin.org




understood what the sounds were, as evidenced by their high ratesof accuracy on the task.) A rater was present in the room recordingeach response on a laptop. Therefore, each stimulus was verballyidentified by the participant immediately after stimulus delivery.As before, the seven animal calls were randomly delivered through-out the block. This second control variant of the WHAT task wasalways run on a separate day. In our analyses, WHAT II will becompared to WHAT I through generation of a statistical clusterplot analysis to determine whether the two paradigms yield signif-icant differences at any electrodes across all time points. WHAT IIwill also be compared to WHERE to determine whether a similarpattern of results is seen between WHERE and the two respectiveWHAT conditions.

DATA ACQUISITIONContinuous EEG was acquired through the ActiveTwo Biosemi™

electrode system from 168 scalp electrodes, digitized at 512 Hz.For display purposes, data were filtered with a low-pass 0-phase shift 96 dB 40 Hz filter after acquisition. The referenceelectrode was assigned in software after acquisition. BioSemireplaces the “ground” electrodes that are used in conventional sys-tems with two separate electrodes: common mode sense (CMS)active electrode and driven right leg (DRL) passive electrode.These two electrodes form a feedback loop, thus rendering them“references.” For a detailed description of the referencing andgrounding conventions used by the Biosemi active electrodesystem, the interested reader is referred to the following web-site: http://www.biosemi.com/faq/cms&drl.htm. All data were re-referenced to the nasion after acquisition, for analysis (in one sub-ject, the supranasion was used for re-referencing, as the nasion wascontaminated by artifact). After each recording session, before theelectrode cap was removed from the subject’s head, the 3D coor-dinates of all 168 electrodes with reference to anatomic landmarkson the head (nasion, pre-auricular notches) were digitized using aPolhemus Magnetic 3D digitizer. EEG was averaged offline. Datawere epoched (−100 ms pre-stimulus to 500 ms post-stimulus)and then averaged. Baseline was defined as the mean voltage over−100 to 0 ms preceding the onset of the stimulus. Trials with blinksand large eye-movements were rejected offline on the basis ofhorizontal and vertical electro-oculogram recordings. An artifactrejection criterion of ±100 μV was used at all other electrode sitesto exclude periods of muscle artifact and other noise-transients.From the remaining artifact-free trials, averages were computed foreach subject. These averages were then visually inspected for eachindividual to ensure that clean recordings with sufficient numbersof trials were obtained and that no artifacts were still included.Across both WHAT and WHERE conditions, the average numberof accepted sweeps was over 2000 trials. Prior to group averag-ing, data at electrodes contaminated by artifacts were interpolatedfor each subject. As described by Greischar et al (2004), sphericalspline interpolation represents a method for increasing power atelectrode sites where data has been contaminated. Data were re-baselined over the interval −50 to 20 ms after interpolating. Datawere ultimately averaged across all subjects (grand mean averages)for visual comparison at the group level and for display purposes.The reader should note that throughout this paper, we use thefamiliar nomenclature of the modified 10–20-electrode system to

refer to the positioning of electrode sites. Since our montage con-tains considerably more scalp-sites than this nomenclature allowsfor, in some cases, we will be referring to the nearest neighboringsite within the 10–20 system.

ANALYSIS STRATEGYFor each electrode, the data for all non-target trials from all sub-jects were collapsed into a single average waveform for each ofthe sound sources and averaged over all seven locations. Thesegroup-averaged waveforms were then visually inspected across allscalp-sites. To constrain our analyses, we first identified the well-characterized auditory components prior to and including the N1:P20, Pa, P1, and N1. Components were identified on the basisof their expected latencies, over scalp regions previously iden-tified as those areas of maximal amplitude for auditory evokedcomponentry (see, for example, Picton et al., 1974; Leavitt et al.,2007). Note that the latencies and topographies of the basic AEPcomponents reported here were entirely typical of those previ-ously reported. Analyses were performed across three pairs ofscalp electrodes for each component; we chose three electrodesat homologous locations on each side of the scalp (for a total ofsix electrodes) that best represented the maximal amplitude ofthe component of interest in a given analysis and averaged acrossthe three electrodes. Statistical analyses were conducted with theSPSS software package (SPSS version 11.5). Limiting our analysesto those electrodes where the components are maximal representsa conservative approach to the analysis of high-density ERP dataand raises the likelihood of missed effects (so-called Type II errors)in these rich datasets. Therefore, an exploratory (post hoc) analy-sis phase was also undertaken (described below under StatisticalCluster Plots).

For each component of interest (P20, Pa, P1, N1), we calcu-lated the area under the waveform for an epoch centered on thepeak of the grand mean (the time window used for each epochwas either 10 or 20 ms depending on the width of the respec-tive component’s peak). We then averaged these measures acrossthe three representative electrode sites for each hemisphere. Thesearea measures were then used as the dependent variable. To inves-tigate differences between the AEPs of subjects in the WHAT andWHERE conditions, we tested each identified component with a2 × 2 analysis of variance (ANOVA). The factors were condition(WHAT versus WHERE) and hemi-scalp (right versus left).

To verify that our object identification task was effective, wecompared the average waveform for responses to targets to thesame for the standard (i.e., non-target) stimuli. The Nd compo-nent (also referred to as the response negativity) has been shown tobe elicited when subjects attend to relevant auditory stimuli, whenrelevance is defined along a physical dimension (Hansen and Hill-yard, 1980; Molholm et al., 2004). The presence of a robust Ndcomponent would provide confirmation that our subjects were infact actively engaged in the object detection task.

It is important to note that our first level of analysis was per-formed using an average of all seven sound sources, collapsed intoa single waveform for each condition. However, the experimentwas designed such that every subject was exposed to hundreds oftrials from each individual sound source, therefore enabling us toanalyze the ERPs generated in response to each spatial location


www.frontiersin.org




separately. As a second level of analysis, we inspected seven sepa-rate waveforms within each condition. We compared these sevenaverages for our subjects both within and across condition.

SOURCE RECONSTRUCTIONSource reconstructions were generated by applying the minimum-norm least squares method as implemented in the BESA softwaresuite (version 5.1; Gräfelfing, Germany). The minimum-normapproach is a commonly used method to estimate a distributedelectrical current image in cortex as a function of time (see e.g.,Hamalainen and Ilmoniemi, 1994). The source activities of 1426regional sources are computed. Since the number of sources ismuch larger than the number of sensors, the inverse problem ishighly underspecified and must be stabilized by a mathematicalconstraint (i.e., the minimum norm). Out of the many currentdistributions that can account for the recorded sensor data, thesolution with the minimum L2 norm, i.e. the minimum totalpower of the current distribution, is displayed in BESA. To com-pute the minimum norm here, we utilized an idealized three-shellspherical head model with a radius of 85 mm and assumed scalpand skull thickness of 6 and 7 mm, respectively. The minimumnorm was applied to the data across the latency interval 90–130 ms,spanning the N1 timeframe. Baseline period was used to computemean noise levels. Weightings were applied to the data as follows:(1) weighting at the individual channel level with a noise scalefactor of 1.00; (2) depth-weighting such that the lead field of eachregional source was scaled with the largest singular value of thesingular value decomposition of the source’s lead field; and (3)spatio-temporal weighting using the signal subspace correlationmethod of Mosher and Leahy (1998) (dimension setting = 6).

STATISTICAL CLUSTER PLOTSAs described above, we took a conservative approach to the analy-sis of the high-density ERP data in order to limit the number ofstatistical tests performed, with the spatio-temporal properties ofthe componentry delimiting the tests. Our conservative approachraises the likelihood of missed effects. We therefore performedan exploratory analysis as a means of fully exploring the rich-ness of our data set and as a hypothesis-generating tool for futureresearch. We have devised a simple method for testing the entiredata matrix for possible effects, which we term statistical clusterplots (see Molholm et al., 2002; Murray et al., 2002). These clus-ter plots were derived by calculating point-wise, paired, two-tailedt -tests between the WHAT and WHERE conditions. The resultswere then arrayed on a single grid, with scalp regions (electrodepositions) plotted on the y axis and post-stimulus time plottedon the x axis, thus providing a snapshot overview of significantdifferences between conditions across scalp regions over time. Inthe present data treatment, periods of significant difference wereonly plotted if an alpha criterion of 0.05 was exceeded and thenonly if this criterion was exceeded for at least five consecutive datapoints (∼10 ms; see Wetherill and Levitt, 1965; Foxe and Simpson,2002).

TOPOGRAPHIC MODULATIONSTo statistically identify periods of topographic modulation, wecalculate the global dissimilarity (GD; Lehmann and Skrandies,

1980, 1984) between WHAT and WHERE responses for eachtime point of each subject’s data. GD is an index of configura-tion differences between two electric fields, independent of theirstrength. This parameter equals the square root of the mean of thesquared differences between the potentials measured at each elec-trode (versus the average reference), each of which is first scaledto unitary strength by dividing by the instantaneous global fieldpower (GFP). GD can range from 0 to 2, where 0 indicates topo-graphic homogeneity and 2 indicates topographic inversion. Toderive statistical significance, the observed GD score at each timepoint was compared against a Monte Carlo distribution composedof 5000 iterations (see Strik et al., 1998 for a description). Statisti-cal significance was determined when the observed GD score (i.e.,the mean GD across participants) was ±2 SD away from the meanof the Monte Carlo distribution. Since electric field changes areindicative of changes in the underlying generator configuration(Fender, 1987; Lehmann, 1987, this non-parametric test, in com-bination with the scalp distributions, provides a reliable measurefor determining if and when the brain networks activated by theWHAT and WHERE conditions differ significantly.

RESULTSBEHAVIORAL RESULTSWHAT condition: group-averaged accuracy across all trials for thesound identification task was 85.4%. WHERE condition: for thesound localization task, 11 participants’ responses were analyzed(one subject’s responses had to be excluded from the behavioralanalysis as they were not captured due to a computer malfunction).Each participant’s responses were averaged separately for each ofthe seven sound source locations; group-averaged responses aredisplayed in Figure 4. The average offset of responses collapsedover all seven locations was 10.6˚ ± 12.0˚. The average offset tosound source 4, the midline sound source (where we might wellexpect a very high degree of accuracy) was 4.5˚; moreover, all 11subjects erred to the left of center. Greater accuracy was seen forresponses to sound sources located in participants’ left hemifield(Sound sources 1, 2, 3), as compared to right hemifield (Soundsources 5, 6, 7). The average offset for responses to left soundsources (averaging over sound sources 1, 2, and 3) was 3.6˚ ± 11.0˚.Average offset to sound sources in subjects’ right hemifield (aver-aging over 5, 6, and 7) was 11.7˚ ± 9.3˚. This between hemifielddifference was significant (p = 0.006).

ELECTROPHYSIOLOGICAL RESULTS I: GENERAL DESCRIPTIONIn both tasks, stimuli elicited clear ERPs, which contained P20,Pa, P1, and N1 components (Figure 5) with typical scalp dis-tributions (Figure 6). The time period 15–25 ms was selectedaround the P20 peak in our dataset. An ANOVA revealed no sig-nificant main effect of task, WHAT versus WHERE, (F 1,11 = 0.006,p = 0.938), nor was there a significant interaction effect for con-dition by hemisphere (F 1,11 = 2.509, p = 0.141). The Pa peak inour data was at approximately 42 ms; analysis of 10-ms time win-dow around the peak of the Pa found neither a main effect ofcondition (F 1,11 = 1.061, p = 0.325), nor for condition by hemi-sphere (F 1,11 = 0.079, p = 0.783). The P1 peak was found atapproximately 62 ms. Analysis of a 10-ms time window aroundthe peak of the P1 yielded neither a main effect of condition


www.frontiersin.org




FIGURE 4 | Behavioral results for the “WHERE” condition. Blue lines identify the average of responses across 11 participants (one had to be excluded due toa technical problem capturing responses) for each of the seven sound source locations. Average offset of responses for all seven locations collapsed was10.6˚ ± 12˚.

(F 1,11 = 0.966, p = 0.347), nor a condition by hemisphere interac-tion (F 1,11 = 0.298, p = 0.596). In our data, the N1 was evident as anegativity that peaked fronto-centrally at approximately 124 ms inboth conditions. To account for the broader deflection of this com-ponent, a 20-ms time window was selected around the peak of theN1 (114–134 ms). Here, we found a significant main effect of con-dition (F 1,11 = 5.811, p = 0.035), reflecting substantially greateramplitude of the N1 in the WHAT condition. The effect size ofthis difference was 0.64 (using the criteria of Cohen’s d). Therewas no condition by hemisphere interaction for the N1 component(F 1,11 = 0.455, p = 0.514).

To rule out the possibility of order effects, subjects who com-pleted the WHERE condition on Day 1 were compared to thosewho completed the WHAT condition on Day 1 (Figure 5B). Asdescribed in the text, the figure demonstrates that both groups(small N of 6 notwithstanding) showed an N1 difference betweenconditions that is consistent with that shown by the full group,and is detailed in our main results for the full group. Importantly,the direction of the difference is consistent with the full groupanalysis.

To ensure that subjects’ responses actually reflected attentionalprocessing of relevant auditory stimuli, we compared the responseevoked by target stimuli with that of non-target stimuli. As shownin Figure 7, the response elicited by stimuli that included a targetauditory element became more negative-going than the responseselicited by the stimuli without a target auditory element overcentral/fronto-central scalp. This difference extended to about350 ms. This response pattern is consistent with elicitation of theauditory selective attention component, the Nd (see, for example,Hansen and Hillyard, 1980; Molholm et al., 2004), thereby pro-viding us with compelling evidence that our subjects were trulyengaged in a task of sound object recognition.

To better detect missed effects between conditions, we com-puted a statistical cluster plot for all seven sound sources collapsedinto one waveform for each condition (see Materials and Meth-ods). This served as a hypothesis-generating tool; it provided uswith a snapshot of where and when significant differences occurredbetween conditions that may have been missed in our plannedanalyses. Consistent with our previous analysis, we observed a dis-tinct cluster that coincided with the N1 (Figure 8). What provedto be rather striking was the almost complete absence of any otherdifferences, particularly at later time-points where higher-ordercognitive effects generally tend to become evident in ERPs.

A topographic map of the differences between conditions forthe N1 timeframe (Figure 9) reveals a topography that is not con-sistent with auditory components generated in primary auditorycortex. Indeed, it is suggestive of substantial contributions fromfrontal generators, thereby reflecting processes that more likelyinvolve ventral auditory pathway.

ELECTROPHYSIOLOGICAL RESULTS II: LOCATION SPECIFIC AVERAGESWe examined each of the seven separate sound source averageswithin each condition to determine whether evidence of an inter-action between sound source location and task could be seen.When we compared between tasks, the N1 difference was consis-tent; that is, the N1 amplitude in the WHAT task was consistentlygreater than in the WHERE task (Figure 10A). To determine thesignificance of this and to probe for any further effects, individualstatistical cluster plots for each sound source location were gener-ated. A significant difference between conditions emerged aroundthe time window for the N1 at each sound source location withthe exception of Sound source 4 (the midline sound source). It isimportant to note that the latency of the difference was shifted,depending on the spatial orientation of the sound source. That is,


www.frontiersin.org




FIGURE 5 | (A) Auditory evoked potentials in participants generated duringthe WHAT task (red) and the WHERE task (black). Data from eight electrodesspanning frontal, central, and centro-parietal regions are presented. (B)

Representative electrode from subgroups of the six subjects who participatedin the WHAT condition first (Day 1) and the six who participated in theWHERE condition first.

for Sound source 1, the between-condition difference emerged theearliest, prior to 100 ms (∼80 ms), while the difference at Soundsource 7 became significant after 100 ms (∼120 ms). A closerinspection of the individual sound source waveforms revealed thatthe N1 peak was actually later for responses to Sound source 7 thanSound source 1 in both conditions, by approximately 10 ms. Thisdetermination was made by a visual inspection of midline elec-trodes where the N1 was maximal; therefore, the effect cannot bedue to laterality effects. For both the WHAT and WHERE tasks,

the N1 responses to each individual sound source location fromone representative electrode where the N1 component is maximal(Fz) were overlaid and these are presented in Figure 10B. Whereas,in the object task, the N1 response elicited to each sound sourcewas highly stable in that the amplitude was very similar acrosssound sources, the spatial task evoked responses to each soundsource location that showed greater variability. To test this statisti-cally, a 2 × 2 × 7 ANOVA was conducted with factors of condition(WHAT versus WHERE), sound source (1–7), and sound source


www.frontiersin.org




FIGURE 6 |Topographic maps of ERPs recorded during WHAT and WHERE tasks. Group-averaged data are displayed for peak of each auditory component:P20, Pa, P1, and N1.

FIGURE 7 | Group-averaged waveforms for responses to the sound

object recognition (WHAT ) task for target stimuli (red) versus

non-target stimuli (blue). Representative electrode shown is FCz.

hemifield (left versus right) to determine whether there was a sig-nificant interaction between sound source location and task. Wetook the average N1 amplitude from three fronto-central elec-trodes (around FCz) using a 40-ms time window around the N1peak (importantly, this window was shifted such that it centeredon the peak of the N1 for each sound source, as these latenciesvaried). No significant interaction was found for sound sourceby condition [F(2,10) = 1.057, p > 0.1], nor was there a three-way interaction between hemifield, sound source, and condition[F(2,10) = 0.410, p > 0.1].

Topographic maps of the difference between conditions(WHAT minus WHERE) were created for an average of the left

hemifield sound sources (1–2–3) and the right hemifield soundsources (5–6–7). Subtle but distinct differences can be seen forthese, suggesting contributions from unique underlying genera-tor configurations to stimuli that come from the left hemifieldcompared to those from the right (Figure 10C).

TOPOGRAPHIC MODULATION ANALYSISAveraging across all seven spatial locations, the GD metric allowedus to identify a distinct period of statistically significant differencesin topographic modulations between the WHAT and WHEREconditions over the N1 period, indicative of the activation ofdistinct configurations of intracranial brain generators for eachexperimental condition (Figure 11). That is, the observed effect atN1 was not only due to a change in the amplitude of activationwithin a given generator configuration, but crucially, to actual dif-ferences in the configurations of neural generators underlying thetwo tasks (either in the actual neural generators involved, or in therelative contributions from the same set of neural generators).

CORTICAL ACTIVATION: MINIMUM-NORM MAPSAs described in the Methods, the baseline period was used tocompute mean noise levels for the source reconstructions, result-ing in a signal-to-noise ratio for the selected epoch of 8.67.Source reconstructions at 10 ms intervals, starting from 90 ms,suggested a strong left hemisphere bias for differential process-ing between conditions (see Figure 12). Two stable clusters ofactivation, suggestive of separable processing streams, can beseen. At 90 ms, the strongest activation is seen in a clusterover left inferior and superior parietal areas, and this activa-tion continues to be apparent throughout the sampled epoch,


www.frontiersin.org




FIGURE 8 | Statistical cluster plot of the results of the point-wise

running two-tailed t -tests comparing the amplitudes of participants’

auditory evoked potentials in the WHAT versus WHERE conditions.

Time with respect to stimulus onset is presented on the x axis and

topographic regions of 168 electrode positions on the y axis. Colorcorresponds to t values. Periods of significant difference are only plottedif a strict alpha criterion of <0.05 was exceeded for at least five consecutivedata points.

FIGURE 9 |Topographic map of the differences between conditions

(WHAT minus WHERE ) during the N1 timeframe.

although it weakens as time progresses. This cluster is anatom-ically compatible with the dorsal auditory pathway. The sec-ond stable cluster, compatible with the ventral auditory path-way, emerges at approximately 100 ms over anterior temporalregions and peaks at 110 ms, after which point it dissipatessubstantially. As such, these two activation clusters appear tohave some WHAT different time courses. Activation in theright hemisphere suggests much less involvement of underlyinggenerators.

RESULTS FOR ALTERNATE WHAT PARADIGM: WHAT IIGiven that only a subset of participants (n = 6) completed theWHAT II paradigm, results should be interpreted cautiously.

However, a very specific prediction was being tested here, namely,that a manipulation of the task would not affect the N1 differ-ence seen between the WHAT and WHERE conditions. Behavioralperformance accuracy in the WHAT II condition did not differsignificantly from WHAT I. A statistical cluster plot analysis com-paring the two WHAT paradigms across the entire epoch (−100to 500) failed to find any areas of significance, suggesting that thedifferences between WHAT I and WHAT II were negligible (plotsnot shown). These results support the notion that one paradigmwas just as effective as the other in characterizing neural processesunderlying sound identification. Next, we conducted an analysiscomparing WHAT II and WHERE. We reasoned that if differencesbetween these tasks were found it would support the view that bothWHAT I and WHAT II are tapping processes of sound object iden-tification, distinct from processes of sound localization. We useda one-tailed t -test, guided by our hypothesis that the differencebetween these conditions would replicate the direction seen forWHAT I versus WHERE. The N1 difference was apparent betweenconditions. Comparing the waveforms, we saw that the directionof the effect was the same: WHAT II resulted in a greater deflec-tion of the N1 relative to WHERE, as seen in WHAT I (Figure 13).While the absolute magnitude of the difference between WHAT IIand WHERE is not the same as that of WHAT I versus WHERE,given that the number of subjects in WHAT II was only a subset ofthe full sample (n = 6), this was likely a power issue and as such,not an unexpected finding.

DISCUSSIONThe present data showed a robust difference in processing overa surprisingly delimited time frame for subjects performing atask of sound object recognition versus sound localization usingexactly the same stimulus setup. This difference was characterizedby greater negative-going amplitude during the N1 processingtimeframe for the WHAT condition compared to the WHERE


www.frontiersin.org




FIGURE 10 | (A) WHAT (red) and WHERE (blue) group-averagedwaveforms taken from electrode Cz are shown for seven individual soundsource averages. In the figure, each waveform pair is shown at a locationconsistent with the sound source from which it was derived. (B) Separatewaveforms for each of the sound sources are overlaid from each conditionat one representative electrode (Fz). (C) Topographic maps of the N1difference between conditions (WHAT minus WHERE ) showing distincttopographies for sound sources on the left (Sound sources 1–2–3)compared to the right (Sound sources 5–6–7).

condition. This effect was seen consistently at electrode sites overfrontal, central, centro-parietal and parietal scalp regions. Whilethis difference occurred during the N1 timeframe, it was clearlynot a simple modulation of N1 amplitude as it had a completelydifferent topographic distribution to that of the N1. GD measures

FIGURE 11 | Statistical plot of topographic modulations. Areas ofnon-significance appear in blue; areas where significance exceeds p < 0.03are displayed in a graduated color scheme according to the legend shown.Note that this significance cluster falls in a time frame coincident with theN1 component.

using topographic modulation analysis confirmed that the differ-ence between tasks during this timeframe was driven by a shiftin the underlying generator configuration. Additionally, applyinga minimum-norm source reconstruction algorithm to the differ-ence wave revealed distinct activations that corresponded well withactivity within putative dorsal and ventral auditory structures.

The main goal of the present investigation was to find the earli-est discernible point of differential processing in the context ofspatial and non-spatial auditory tasks. However, a crucial dis-tinction must be made here, and that is that feature processingpathways in cortex are not laid out as one-way highways. Namely,when auditory information is received, it comprises both spatialand non-spatial object features. While the organism may be biasedto preferentially process one or the other of these aspects at anygiven time, this is not to say that the stimulus is stripped of itsunattended features. Findings from studies of MMN make clearthat both location and object-based properties of auditory stimuliare automatically processed even when attention is directed awayfrom the auditory modality (e.g., Paavilainen et al., 1989; Molholmet al., 2005; Tata and Ward, 2005; Ritter et al., 2006; Näätänen et al.,2007); indeed, MMNs to these features are automatically elicitedin people when they are asleep (e.g., Ruby et al., 2008) or evenin a coma (e.g., Fischer et al., 2000; Wijnen et al., 2007). More-over, there is a clear suggestion from our prior investigation ofmid-latency auditory evoked responses that there is a constitutivedivergence of activation into separable pathways in the context ofa passive listening task (Leavitt et al., 2007). While that study wasnot designed to expressly test the functionality of the dorsal andventral auditory pathways, automatic sensory activation of thesepathways is an inherent result of receiving any auditory stimula-tion or performing any auditory task. Here, we asked subjects toactively attend to one stimulus feature, location, or identity, in agiven condition. Clearly, however, neural processing of the “unat-tended”feature occurred to some degree as well. There is a growingbody of functional imaging literature showing that irrelevant fea-tures of objects seem to be processed by the brain, despite havingno current task relevance (e.g., O’Craven et al., 1999; Wylie et al.,2004). This strong bias for attention to spread to non-primary


www.frontiersin.org




FIGURE 12 | Cortical maps displaying the spatio-temporal minimum-norm topographies for WHAT minus WHERE difference waveforms.

FIGURE 13 | Waveforms from participants (n = 6) comparing performance

in WHAT II and WHERE conditions. The direction of the N1 amplitudedifference between conditions mirrors that seen in WHAT I versus WHERE ;that is, the N1 amplitude for WHAT II is greater than WHERE.

stimulus features has also been shown cross-modally, using audio-visual tasks (Molholm et al., 2004, 2007; Fiebelkorn et al., 2010a,b).This spread of attention is further evidenced by findings in thevisual modality that directing spatial attention to one part of anobject results in the facilitation of sensory processing across theentire extent of that object (e.g., Egly et al., 1994; Martinez et al.,2007). Therefore, while in the present investigation we have iden-tified a point of task-specific divergence at approximately 120 ms,it is extremely unlikely that this represents the first point at whichprocessing for these qualities truly begins.

Two previous studies of object versus spatial processing havealso shown amplitude modulation during the auditory N1 time-frame, but in both of these, the direction of the effect was actually

opposite to that seen here, with responses in the WHERE condi-tion showing larger amplitude (Ahveninen et al., 2006; De Santiset al., 2007a). A third MEG study found no amplitude differencesfor N1 (Anourova et al., 2001), but rather, a latency differencebetween tasks. These differences are likely explained by paradig-matic differences. The stimuli used by these investigators weresimple tones (Anourova et al., 2001; De Santis et al., 2007a) andvowel pairs (Ahveninen et al., 2006) and it could be argued thatthese stimuli might be relatively less effective at activating object-processing regions. Here, by employing complex, and considerablymore ecological sounds, it is likely that greater activation of WHATprocessing regions was invoked. Whereas sound localization maybe more automatic, and engaged more quickly (as suggested by


www.frontiersin.org




Anourova et al., 2001), object recognition and identification reliesmore heavily on accessing a memory trace and it is a reasonableassumption that invoking a memory trace of a complex soundsuch as the animal calls used here, would employ a greater neuralnetwork than that of a simple tone. Similarly, in all of these priorinvestigations, only a limited number of spatial locations wereused during the spatial condition. In most, only two positionswere used, and in only one case were three used (Anurova et al.,2003). Here, seven discrete locations were used, and therefore per-forming the current task would have required a great deal moreeffort, and consequently, greater involvement of underlying neuralnetworks for spatial discrimination. Moreover, our subjects werenot explicitly aware that there were seven discrete sound sources,and so the number of sound sources was unknown to them andlikely created higher demands on them relative to past studies uti-lizing less sound sources. In sum, we suggest that the paradigmemployed in the current study was both: (1) more challengingin both the WHAT and WHERE conditions than prior studiesemploying less complex, and fewer stimulus types, and (2) moreecologically valid, and therefore provided results which more accu-rately reflect the way we actually process auditory information inthe complex world around us.

It could be considered surprising that while we found signifi-cant task-related differences during the N1 processing timeframe,there were no subsequent time-points where significant differ-ences were seen. With such different tasks as localizing sound andmaking a sound object identification, one might well predict thatdistinct higher-order cognitive processes might be engaged at laterlatencies, such as the P300 component, which has been shownto index attention-related information processing (e.g., Polich,1986; Linden, 2005). On the other hand, the lack of differencesduring later timeframes here provides good confirmation thathigher-order attentional processes were not differentially engagedbetween the two tasks, suggesting that the location and recognitiontasks were well-matched for difficulty. This suggestion, however,must be considered in the context of our notion that the use of ani-mal calls in the present study is what made our design both moreecologically valid, and more accurate in identifying a greater allo-cation of neural resources as evidenced exclusively by the higheramplitude N1 component seen in the WHAT condition. Whethersuch a focal deployment of attentional resources (that is, limitedonly to the N1 period) can be measured in an experimental para-digm engaging sensory modalities other than the auditory systemremains to be investigated by future studies.

The behavioral findings for the WHERE task are noteworthy,in that they appear to represent an auditory analog to the patternof pseudoneglect that has been consistently observed in the visualsystem of healthy controls. Specifically, pseudoneglect refers to aleftward attentional bias that occurs in normally functioning indi-viduals and causes them to bisect visually presented lines slightlyto the left of their true center (e.g., McCourt, 2001; Foxe et al.,2003; McCourt et al., 2008). This spatial bias has been attrib-uted to the well-known right hemisphere specialization for spatialattention (see e.g., Mesulam, 2000), and indeed, both functionalimaging (Fink et al., 2001 and ERP studies (Foxe et al., 2003) showthat visuospatial judgments during line-bisection are subserved byright parietal regions. Here, subjects consistently mislocated the

centrally presented sound approximately 4.5˚ leftward of veridicalcenter. This leftward bias can also be seen for the pair of closestflanking sound sources (sound sources 3 and 5). Another interest-ing aspect of these data was the observation of significantly moreaccurate responses to the three sound sources in the left hemi-sphere (Sound sources 1, 2, 3) than to the three sound sourcesin the right hemisphere (Sound sources 5, 6, 7). This is consis-tent with what is known about hemispheric differences in soundlocalization in humans; namely, that auditory spatial processing ismore accurate for sounds in the left hemifield (Burke et al., 1994),again presumably because of the right hemisphere bias for spa-tial attention. Indeed, these investigators interpreted their findingsto suggest a pivotal role for right hemisphere in auditory spatialacuity. In agreement, other studies have implicated greater involve-ment of the right hemisphere in sound localization (see e.g., Farahet al., 1989; Hausmann et al., 2005; Tiitinen et al., 2006; Mathiaket al., 2007).

It is of note that we did not find specific lateralization of themain effect reported here between the WHAT and WHERE tasks,with no evidence for a hemisphere by condition interaction in ourmain ERP analysis. However, the ensuing source-localization didsuggest that there may in fact be a left hemisphere bias, whichwould be in accord with literature providing evidence for a leftlateralization effect of the WHERE pathway during speech soundprocessing (Mathiak et al., 2007), as well as evidence for involve-ment of language-dominant inferior dorsolateral frontal lobe inprocessing non-verbal auditory information (Mathiak et al., 2004).It will fall to future work to specifically test whether there ishemispheric specialization for processing of WHAT and WHEREauditory stimuli.

HUMAN LESION STUDIESHuman lesion studies provide additional support for the divisionof the auditory system into dorsal and ventral pathways (Clarkeet al., 2000, 2002; Clarke and Thiran, 2004). In a series of case stud-ies, Clarke et al. (2000) investigated auditory recognition and local-ization in four patients with circumscribed left hemisphere lesions.Two of four patients showed impaired recognition of environmen-tal sounds but intact ability to localize sounds. In one, the lesionincluded postero-inferior portions of the frontal convexity and theanterior third of the temporal lobe. In the other, the lesion includedleft superior, middle and inferior temporal gyri, and lateral audi-tory areas, but spared Heschl’s gyrus, the acoustic radiation, andthe thalamus. The third patient was impaired in auditory motionperception and localization, but had preserved environmentalsound recognition; here, the lesion included the parieto-frontalconvexity and the supratemporal region. The fourth patient wasimpaired in both tasks, with a lesion comprising large portions ofthe supratemporal region, temporal, postero-inferior frontal, andantero-inferior parietal convexities. A subsequent investigation bythe same group examined 15 patients with right focal hemisphericlesions (Clarke et al., 2002). Again, sound recognition and soundlocalization were shown to be disrupted independently. In thisstudy, patients with selective sound localization deficits tendedto have lesions involving inferior parietal and frontal cortices, aswell as superior temporal gyrus, whereas patients with selectiverecognition deficits had lesions of the temporal pole, the anterior


www.frontiersin.org




part of fusiform, and the inferior and middle temporal gyrus. Thisdouble dissociation clearly supports somewhat independent pro-cessing streams for sound recognition and sound localization inhumans.

CONCLUSIONIn this study, we asked participants to make judgments aboutthe spatial location of sounds versus the object-identity of thosesame sounds, with an eye to assessing whether varying task set inthis manner would engage dissociable neural circuits within audi-tory cortices. These tasks were chosen to tap so-called WHEREand WHAT processes, which have been associated with dor-sal and ventral regions of the auditory system respectively. Ourhigh-density electrical mapping data revealed a robust differencein the ERP during the timeframe of the auditory N1 component

and topographic analysis pointed to differential engagementof underlying auditory cortices during this timeframe. Source-analysis supported the main thesis in that this differential process-ing was generated in dissociable regions of the dorsal and ventralauditory processing stream.

ACKNOWLEDGMENTSSupport for this work was provided by a grant from the USNational Institute of Mental Health (RO1-MH65350 to John J.Foxe; RO1-MH85322 to Sophie Molholm and John J. Foxe) andby a Ruth L. Kirschstein pre-doctoral fellowship to Victoria M.Leavitt (NRSA – MH74284). The authors thank Daniella Blanco,Megan Perrin, and Jennifer Montesi for their invaluable assistancewith data collection.

REFERENCESAhveninen, J., Jaaskelainen, I. P., Raij,

T., Bonmassar, G., Devore, S.,Hamalainen, M., Levanen, S., Lin, F.H., Sams, M., Shinn-Cunningham,B. G., Witzel, T., and Belliveau, J. W.(2006). Task-modulated “what” and“where” pathways in human audi-tory cortex. Proc. Natl. Acad. Sci.U.S.A. 103, 14608–14613.

Alain, C., Arnott, S. R., Hevenor, S., Gra-ham, S., and Grady, C. L. (2001).“WHAT” and “where” in the humanauditory system. Proc. Natl. Acad.Sci. U.S.A. 98, 12301–12306.

Anourova, I., Nikouline, V. V.,Ilmoniemi, R. J., Hotta, J., Aro-nen, H. J., and Carlson, S. (2001).Evidence for dissociation of spatialand nonspatial auditory informa-tion processing. Neuroimage 14,1268–1277.

Anurova, I., Artchakov, D., Korvenoja,A., Ilmoniemi, R. J., Aronen, H. J.,and Carlson, S. (2003). Differencesbetween auditory evoked responsesrecorded during spatial and non-spatial working memory tasks. Neu-roimage 20, 1181–1192.

Arnott, S. R., Binns, M. A., Grady, C.L., and Alain, C. (2004). Assessingthe auditory dual-pathway model inhumans. Neuroimage 22, 401–408.

Belin, P., and Zatorre, R. J. (2000).“WHAT”, “where” and “how” inauditory cortex. Nat. Neurosci. 3,965–966.

Burke, K. A., Letsos, A., and Butler,R. A. (1994). Asymmetric perfor-mances in binaural localization ofsound in space. Neuropsychologia 32,1409–1417.

Clarke, S., Bellmann, A., Meuli, R.A., Assal, G., and Steck, A. J.(2000). Auditory agnosia and audi-tory spatial deficits following lefthemispheric lesions: evidence fordistinct processing pathways. Neu-ropsychologia 38, 797–807.

Clarke, S., Bellmann Thiran, A., Maeder,P., Adriani, M., Vernet, O., Regli,L., Cuisenaire, O., and Thiran, J. P.(2002). WHAT and where in humanaudition: selective deficits followingfocal hemispheric lesions. Exp. BrainRes. 147, 8–15.

Clarke, S., and Thiran, A. B. (2004).Auditory neglect: WHAT and wherein auditory space. Cortex 40,291–300.

De Santis, L., Clarke, S., and Murray, M.M. (2007a). Automatic and intrinsicauditory “WHAT” and “where” pro-cessing in humans revealed by elec-trical neuroimaging. Cereb. Cortex17, 9–17.

De Santis, L., Spierer, L., Clarke, S.,and Murray, M. M. (2007b). Get-ting in touch: segregated somatosen-sory WHAT and where pathwaysin humans revealed by electri-cal neuroimaging. Neuroimage 37,890–903.

Egly, R., Driver, J., and Rafal, R.D. (1994). Shifting visual attentionbetween objects and locations: evi-dence from normal and parietallesion subjects. J. Exp. Psychol. Gen.123, 161–177.

Fabiani, M., Kazmerski, V. A., Cycow-icz, Y. M., and Friedman, D. (1996).Naming norms for brief envi-ronmental sounds: effects of ageand dementia. Psychophysiology 33,462–475.

Farah, M. J., Wong, A. B., Monheit,M. A., and Morrow, L. A. (1989).Parietal lobe mechanisms of spa-tial attention: modality-specific orsupramodal? Neuropsychologia 27,461–470.

Fender, D. H. (1987). “Source local-ization of brain electrical activity,”in Handbook of Electroencephalog-raphy and Clinical Neurophysiology,Revised Series, Vol. 1: Methods ofAnalysis of Brain Electrical and Mag-netic Signals, eds A. S. Gevins and

A. Remond (Amsterdam: Elsevier),355—399.

Fiebelkorn, I. C., Foxe, J. J., Schwartz, T.H., and Molholm, S. (2010a). Stay-ing within the lines: the formationof visuospatial boundaries influ-ences multisensory feature integra-tion. Eur. J. Neurosci. 31, 1737–1743.

Fiebelkorn, I. C., Foxe, J. J., and Mol-holm, S. (2010b). Dual mechanismsfor the cross-sensory spread of atten-tion: how much do learned asso-cations matter? Cereb. Cortex 20,109–120.

Fink, G. R., Marshall, J. C., Weiss, P.H., and Zilles, K. (2001). The neuralbasis of vertical and horizontal linebisection judgments: an fMRI studyof normal volunteers. Neuroimage14(1 Pt 2), S59—S67.

Fischer, C., Morlet, D., and Giard, M.(2000). Mismatch negativity andN100 in comatose patients. Audiol.Neurootol. 5, 192–197.

Foxe, J. J., McCourt, M. E., and Javitt,D. C. (2003). Right hemispherecontrol of visuospatial attention:line-bisection judgments evaluatedwith high-density electrical map-ping and source analysis. Neuroim-age 19, 710–726.

Foxe, J. J., and Simpson, G. V. (2002).Flow of activation from V1 to frontalcortex in humans. A framework fordefining “early” visual processing.Exp. Brain Res. 142, 139–150.

Greischar, L. L., Burghy, C. A., vanReekum, C. M., Jackson, D. C., Piz-zagalli, D. A., Mueller, C., and David-son, R. J. (2004). Effects of electrodedensity and electrolyte spreading indense array electroencephalographicrecording. Clin. Neurophysiol. 115,710—720.

Hamalainen, M. S., and Ilmoniemi, R. J.(1994). Interpreting magnetic fieldsof the brain: minimum norm esti-mates. Med. Biol. Eng. Comput. 32,35–42.

Hansen, J. C., and Hillyard, S. A. (1980).Endogenous brain potentials associ-ated with selective auditory atten-tion. Electroencephalogr. Clin. Neu-rophysiol. 49, 277–290.

Hausmann, M., Corballis, M. C., Fabri,M., Paggi, A., and Lewald, J. (2005).Sound lateralization in subjects withcallosotomy, callosal agenesis, orhemispherectomy. Brain Res. Cogn.Brain Res. 25, 537–546.

Kaas, J. H., and Hackett, T. A. (1999).“WHAT” and “where” processing inauditory cortex. Nat. Neurosci. 2,1045–1047.

Lalor, E. C., Yeap, S., Reilly, R. B., Pearl-mutter, B. A., and Foxe, J. J. (2008).Dissecting the cellular contributionsto early visual sensory processingdeficits in schizophrenia using theVESPA evoked response. Schizophr.Res. 98, 256–264.

Leavitt, V. M., Molholm, S., Ritter, W.,Shpaner, M., and Foxe, J. J. (2007).Auditory processing in schizophre-nia during the middle latency period(10–50 ms): high-density electricalmapping and source analysis revealsubcortical antecedents to early cor-tical deficits. J. Psychiatry Neurosci.32, 339–353.

Lehmann, D., and Skrandies, W. (1980).Reference-free identification ofcomponents of checkerboard-evoked multichannel potentialfields. Electroencephalogr. Clin.Neurophysiol. 48, 609–621.

Lehmann, D., and Skrandies, W. (1984).Spatial analysis of evoked potentialsin man – a review. Prog. Neurobiol.23, 227–250.

Lehmann, D. (1987). “Principles of spa-tial analysis,” in Methods of Analysisof Brain Electrical and Magnetic Sig-nals, Handbook of Electroencephalog-raphy and Clinical Neurophysiology,Revised Series, Vol. 1, eds. A. S.Gevins and A. Remond (Amster-dam: Elsevier), 309—354.


www.frontiersin.org




Linden, D. E. (2005). The p300: where inthe brain is it produced and WHATdoes it tell us? Neuroscientist 11,563–576.

Martinez, A., Ramanathan, D. S., Foxe,J. J., Javitt, D. C., and Hillyard, S. A.(2007). The role of spatial attentionin the selection of real and illusoryobjects. J. Neurosci. 27, 7963–7973.

Mathiak, K., Hertrich, I. Grodd, W.,and Ackermann, H. (2004). Dis-crimination of temporal informa-tion at the cerebellum: functionalmagnetic resonance imaging of non-verbal auditory memory. Neuroim-age 21, 154–162.

Mathiak, K., Menning, H., Hertrich,I., Mathiak, K. A., Zvyagintsev, M.,and Ackermann, H. (2007). Who istelling what from where? A func-tional magnetic resonance imagingstudy. Neuroreport 18, 405–409.

McCourt, M. E. (2001). Performanceconsistency of normal observers inforced-choice tachistoscopic visualline bisection. Neuropsychologia 39,1065–1076.

McCourt, M. E., Shpaner, M., Javitt, D.C., and Foxe, J. J. (2008). Hemi-spheric asymmetry and callosal inte-gration of visuospatial attention inschizophrenia: a tachistoscopic linebisection study. Schizophr. Res. 102,189–196.

Mesulam, M. M. (2000). “Attentionalnetworks, confusional states, andneglect syndromes,” in Principles ofBehavioral and Cognitive Neurology,ed. M. M. Mesulam (Oxford: OxfordUniversity Press), 174–256.

Mishkin, M., Lewis, M. E., and Unger-leider, L. G. (1982). Equivalenceof parieto-preoccipital subareas forvisuospatial ability in monkeys.Behav. Brain Res. 6, 41–55.

Molholm, S., Martinez, A., Ritter, W.,Javitt, D. C., and Foxe, J. J. (2005).The neural circuitry of pre-attentiveauditory change-detection: an fMRIstudy of pitch and duration mis-match negativity generators. Cereb.Cortex 15, 545–551.

Molholm, S., Martinez, A., Shpaner,M., and Foxe, J. J. (2007). Object-based attention is multisensory: co-activation of an object’s representa-tions in ignored sensory modalities.Eur. J. Neurosci. 26, 499–509.

Molholm, S., Ritter, W., Javitt, D. C.,and Foxe, J. J. (2004). Multisensoryvisual-auditory object recognitionin humans: a high-density electricalmapping study. Cereb. Cortex 14,452–465.

Molholm, S., Ritter, W., Murray, M.M., Javitt, D. C., Schroeder, C. E.,and Foxe, J. J. (2002). Multisensoryauditory-visual interactions duringearly sensory processing in humans:a high-density electrical mappingstudy. Brain Res. Cogn. Brain Res. 14,115–128.

Mosher, J. C., and Leahy, R. M.(1998). Recursive MUSIC: a frame-work for EEG and MEG sourcelocalization. IEEE Trans. Biomed.Eng. 45, 1342–1354.

Murray, M. M., Wylie, G. R., Higgins,B. A., Javitt, D. C., Schroeder, C. E.,and Foxe, J. J. (2002). The spatiotem-poral dynamics of illusory contourprocessing: combined high-densityelectrical mapping, source analysis,and functional magnetic resonanceimaging. J. Neurosci. 22, 5055–5073.

Näätänen, R., Paavilainen, P., Rinne,T., and Alho, K. (2007). The mis-match negativity (MMN) in basicresearch of central auditory process-ing: a review. Clin. Neurophysiol. 118,2544–2590.

Nousak, J. M., Deacon, D., Ritter, W.,and Vaughan, H. G. Jr. (1996). Stor-age of information in transient audi-tory memory. Brain Res. Cogn. BrainRes. 4, 305–317.

O’Craven, K. M., Downing, P. E., andKanwisher, N. (1999). fMRI evi-dence for objects as the units ofattentional selection. Nature 401,584–587.

Oldfield, R. C. (1971). The assess-ment and analysis of handedness:the Edinburgh inventory. Neuropsy-chologia 9, 97–113.

Paavilainen, P., Karlsson, M. L.,Reinikainen, K., and Näätänen,R. (1989). Mismatch negativityto change in spatial location ofan auditory stimulus. Electroen-cephalogr. Clin. Neurophysiol. 73,129–141.

Picton, T. W., Hillyard, S. A., Krausz,H. I., and Galambos, R. (1974).Human auditory evoked potentials.I. Evaluation of components. Elec-troencephalogr. Clin. Neurophysiol.36, 179–190.

Polich, J. (1986). Attention, probabil-ity, and task demands as deter-minants of P300 latency fromauditory stimuli. Electroencephalogr.Clin. Neurophysiol. 63, 251–259.

Poremba, A., Saunders, R. C., Crane,A. M., Cook, M., Sokoloff, L.,and Mishkin, M. (2003). Functionalmapping of the primate auditorysystem. Science 299, 568–572.

Rämä, P., Poremba, A., Sala, J. B., Yee,L., Malloy, M., Mishkin, M., andCourtney, S. M. (2004). Dissocia-ble functional cortical topographiesfor working memory maintenanceof voice identity and location. Cereb.Cortex 14, 768—780.

Rauschecker, J. P., and Tian, B. (2000).Mechanisms and streams for pro-cessing of “what” and “where” inauditory cortex. Proc. Natl. Acad. Sci.U.S.A. 97, 11800–11806.

Ritter, W., De Sanctis, P., Molholm, S.,Javitt, D. C., and Foxe, J. J. (2006).Preattentively grouped tones do notelicit MMN with respect to eachother. Psychophysiology 43, 423–430.

Roeber, U., Widmann, A., and Schroger,E. (2003). Auditory distraction byduration and location deviants: abehavioral and event-related poten-tial study. Brain Res. Cogn. Brain Res.17, 347–357.

Romanski, L. M., Tian, B., Fritz, J.,Mishkin, M., Goldman-Rakic, P. S.,and Rauschecker, J. P. (1999). Dualstreams of auditory afferents tar-get multiple domains in the primateprefrontal cortex. Nat. Neurosci. 2,1131–1136.

Ruby, P., Caclin,A., Boulet, S., Delpuech,C., and Morlet, D. (2008). Oddsound processing in the sleepingbrain. J. Cogn. Neurosci. 20, 296–311.

Schmolesky, M. T., Wang, Y., Hanes,D. P., Thompson, K. G., Leutgeb,S., Schall, J. D., and Leventhal, A.G. (1998). Signal timing across themacaque visual system. J. Neuro-physiol. 79, 3272–3278.

Schroeder, C. E., Mehta, A. D., andGivre, S. J. (1998). A spatiotempo-ral profile of visual system activa-tion revealed by current source den-sity analysis in the awake macaque.Cereb. Cortex 8, 575–592.

Strik, W. K., Fallgatter, A. J., Bran-deis, D., and Pascual-Marqui, R. D.(1998). Three-dimensional tomog-raphy of event-related potentialsduring response inhibition: evidencefor phasic frontal lobe activation.Electroencephalogr. Clin. Neurophys-iol. 108, 406–413.

Tata, M. S., and Ward, L. M. (2005).Early phase of spatial mismatchnegativity is localized to a poste-rior “where” auditory pathway. Exp.Brain Res. 167, 481–486.

Tian, B., Reser, D., Durham, A., Kus-tov, A., and Rauschecker, J. P. (2001).Functional specialization in rhe-sus monkey auditory cortex. Science292, 290–293.

Tiitinen, H., Salminen, N. H., Palo-maki, K. J., Makinen, V. T., Alku,P., and May, P. J. (2006). Neuro-magnetic recordings reveal the tem-poral dynamics of auditory spatialprocessing in the human cortex.Neurosci. Lett. 396,17–22.

Weeks, R. A., Aziz-Sultan, A., Bushara,K. O., Tian, B., Wessinger, C. M.,Dang, N., Rauschecker, J. P., andHallett, M. (1999). A PET study ofhuman auditory spatial processing.Neurosci. Lett. 262, 155–158.

Wetherill, G. B., and Levitt, H. (1965).Sequential estimation of points on apsychometric function. Br. J. Math.Stat. Psychol. 18, 1–10.

Wijnen, V. J., van Boxtel, G. J., Eilan-der, H. J., and de Gelder, B.(2007). Mismatch negativity pre-dicts recovery from the vegeta-tive state. Clin. Neurophysiol. 118,597–605.

Wylie, G. R., Javitt, D. C., and Foxe, J. J.(2004). Don’t think of a white bear:an fMRI investigation of the effectsof sequential instructional sets oncortical activity in a task-switchingparadigm. Hum. Brain Mapp. 21,279–297.

Zatorre, R. J., Bouffard, M.,Ahad, P., andBelin, P. (2002). Where is “where”in the human auditory cortex? Nat.Neurosci. 5, 905–909.

Conflict of Interest Statement: Theauthors declare that the research wasconducted in the absence of anycommercial or financial relationshipsthat could be construed as a potentialconflict of interest.

Received: 10 August 2010; paper pendingpublished: 17 September 2010; accepted:17 May 2011; published online: 22 June2011.Citation: Leavitt VM, Molholm S,Gomez-Ramirez M and Foxe JJ (2011)“What” and “Where” in auditory sensoryprocessing: a high-density electrical map-ping study of distinct neural processesunderlying sound object recognition andsound localization. Front.Integr. Neurosci. 5:23. doi:10.3389/fnint.2011.00023Copyright © 2011 Leavitt, Molholm,Gomez-Ramirez and Foxe. This is anopen-access article subject to a non-exclusive license between the authors andFrontiers Media SA, which permits use,distribution and reproduction in otherforums, provided the original authors andsource are credited and other Frontiersconditions are complied with.


http://dx.doi.org/10.3389/fnint.2011.00023

www.frontiersin.org



``what'' and ``where'' in auditory sensory processing: a

Documents